Make model preprocessing parallelized across 5 folds

[riyavsinha]

This can mean that there is just a module for running across a single fold, which will be run 5 times, and then one for aggregating the peak distribution across the 5 folds